[Skip to Navigation]
December 1994

The Boundary of Social Phobia: Exploring the Threshold

Author Affiliations

From the Anxiety and Traumatic Stress Program (Dr Davidson), the Departments of Psychiatry (Drs Davidson, George, and Blazer), Medicine (Dr Hughes), and Sociology (Dr George), the School of Nursing (Dr Hughes), and the Office of Medical Education (Dr Blazer), Duke University Medical Center, Durham, NC.

Arch Gen Psychiatry. 1994;51(12):975-983. doi:10.1001/archpsyc.1994.03950120047008

Background:  Individuals with subthreshold social phobia (SSP) in the community are characterized relative to nonphobic, healthy controls (C), and diagnosed social phobics (SP).

Methods:  Data from 1488 subjects from the Duke University Epidemiological Catchment Area Study were examined. Bivariate and multivariate methods were used to compare the SSP, SP, and C groups on 10 sets of variables.

Results:  Compared with C respondents, SSP respondents were more likely to be female and unmarried and to report less income and education. The SSP respondents were also more likely to report work attendance problems, poor grades in school, symptoms of conduct disturbance, impaired subjective social support, lack of self-confidence, lack of a close friend, use of psychotropic drugs in past year, and a greater number of life changes, chronic medical problems, and mental health visits within the past 6 months. In a multivariate logistic regression model with group membership as the dependent variable, compared with C respondents, SSP respondents were more likely to be female, to have less education, and to report more indicators of poor school performance and symptoms of adolescent conduct disturbance. In contrast, in a similar but separate multivariate model, compared with SP respondents, SSP respondents met the criteria for fewer DSM-III psychiatric disorders and were less likely to report impaired instrumental support.

Conclusion:  Social phobia adversely affects over 10% of the population. Previous epidemiologic catchment areabased prevalence estimates have probably been unrealistically low.

Add or change institution